Seongyeop Jeong 정성엽

News

• Dec 2026Upcoming oral presentation at AGU26 on AI prediction of forecast-error trends for operational forecast support.
• Oct 2026Returned to KAIST SAIL to resume my Ph.D. research on explainable AI.
• Oct 2026Poster at the Asian Conference on Meteorology (ACM 2026), Gyeongju, on deep-learning grid-scale correction of radar precipitation.
• Dec 2025Invited talk at the 2025 KAIST XAI Tutorial Series on AI for precipitation forecasting and NWP error estimation.
• Oct 2025Two Korean patents granted (temperature prediction; user-feedback-based AI model customization).
• 2025First-author paper on automated artifact correction in neural TTS published in IEEE TASLP.
• Dec 2023Best Paper Award at the Korea Software Congress (KSC 2023).

Publications

Google Scholar: 74 citations · h-index 2 (as of Oct 2026). Bold = me.

International Journals

1. An Automated Method to Correct Artifacts in Neural Text-to-Speech Models
   Seongyeop Jeong, J. Sung, Inchul Hwang, Jaesik Choi
   IEEE Transactions on Audio, Speech and Language Processing (TASLP), 2025
   JournalFirst authorDOIProject pageCited by 2
2. Temperature Prediction Based on Bidirectional Long Short-Term Memory and Convolutional Neural Network Combining Observed and Numerical Forecast Data
   Seongyeop Jeong, Inyoung Park, Hyun Soo Kim, Chul Han Song, Hong Kook Kim
   Sensors (MDPI), 21(3), 941, 2021
   JournalFirst authorDOICited by 54

International Conferences & Presentations

1. A Deep Learning Approach to Grid-Scale Correction of Radar-Based Precipitation Estimates Using Rain Gauge Observations
   Hyun-Kyoung Lee, Hyun-Kyung Lee, Seongyeop Jeong, Seonghun Pyo, Sangyoun Moon
   Asian Conference on Meteorology (ACM 2026), Gyeongju, Korea
   PosterProgram
2. AI-Based Prediction of Forecast-Error Trends for Operational Forecast Support: Application to 500-hPa Geopotential Height
   MiEun Park, Younggil Choi, Seongyeop Jeong, Soojeong Han, Hyesook Lee
   AGU Annual Meeting 2026 (AGU26), Oral presentation A43F-07, Dec 10, 2026
   OralAbstract
3. Explainable AI-Based Interface System for Weather Forecasting Model
   Soyeon Kim, Jun-Ho Choi, Yejin Choi, Subeen Lee, Artyom Stitsyuk, Minkyoung Park, Seongyeop Jeong, You-Hyun Baek, Jaesik Choi
   HCI International 2023 – Late Breaking Papers, LNCS (Springer)
   ConferenceDOICited by 18
4. The Value of the Onboard Fishing Service and the ICT Technique are Used in 'Ship Matching Service Application'
   Seongyeop Jeong, et al.
   Asian Conference on Engineering Education (ACEE 2018)
   PaperFirst author

Domestic (Korean) Conferences & Workshops

1. Automated Correction of Artifacts in Neural Text-to-Speech Models
   Seongyeop Jeong, Jaesik Choi
   Korea Software Congress (KSC 2023), KIISE, pp. 1092–1094 — Best Paper Award
   PaperFirst authorLink
2. Which Explanation is the Best? A Systematic Evaluation of Feature Attribution Methods on Precipitation Forecasting Model
   Soyeon Kim, Junho Choi, Seongyeop Jeong, Jaesik Choi
   KMS Spring Conference 2023
   Poster
3. A User-Centered Explainable AI Interface System for Weather Forecasting AI Models
   기상예측 인공지능 모델을 위한 사용자 중심 설명가능 인공지능 인터페이스 시스템
   Soyeon Kim, Junho Choi, Subeen Lee, Artyom Stitsyuk, Yeji Choi, Minkyoung Park, Seongyeop Jeong, You-Hyun Baek, Jaesik Choi
   KCC 2023 Workshop on Explainable AI (XAI)
   PosterLink
4. Input-Variable Importance Analysis of a Deep-Learning Precipitation Nowcasting Model Using Explainability and Multi-Modality
   딥러닝 기반 초단기 강수 예측 모델의 설명 가능성과 멀티 모달리티를 활용한 입력 변수 중요도 분석
   Seungheon Shin, Soyeon Kim, Junho Choi, Seongyeop Jeong, Yeji Choi, Jaesik Choi
   Korean Meteorological Society (KMS) Spring Conference 2023
   AbstractLink
5. Inference-Explanation Techniques and Explanation Display Methods for an Explainable-AI-Based Weather Prediction Model
   설명가능한 인공지능기반 기상예측 모델 추론 설명 기술 및 설명 표출방식 개발
   Soyeon Kim, Junho Choi, Seongyeop Jeong, Jaesik Choi
   KMS Spring Conference 2023 — NIMS Outstanding R&D Results Special Session
   OralLink
6. Automatically Correcting Attention-Alignment Errors in Speech Synthesis Models
   음성 합성 모델의 주의 집중 기법 정렬 오류를 자동으로 수정하는 방법
   Seongyeop Jeong, Jaesik Choi
   KCC 2023 Workshop on Explainable AI (XAI)
   PosterFirst authorLink
7. Explainable Artificial Intelligence for Precipitation Forecasting
   Soyeon Kim, Junho Choi, Seongyeop Jeong, Yeji Choi, Jaesik Choi
   KMS Autumn Conference 2022
   AbstractLink
8. An Attention-Based Temperature Prediction Model Using Numerical Weather Prediction and Observation Data
   수치예보모델 자료와 관측 자료를 활용한 주의 집중 기법 기반 온도 예측 모델에 관한 연구
   Seongyeop Jeong, Inyoung Park, Hyun Soo Kim, Chul Han Song, Hong Kook Kim
   Workshop on Image Processing and Image Understanding (IPIU 2020)
   PaperFirst authorLink
9. A CNN-Based Particulate-Matter Prediction Model Using Numerical Weather Prediction Forecasts
   수치예보 모델 예보자료를 활용한 합성곱 신경망 기반 미세먼지 예측모델에 관한 연구
   Seongyeop Jeong, Inyoung Park, Hyun Soo Kim, Chul Han Song, Hong Kook Kim
   29th Joint Conference on Signal Processing (2019)
   PaperFirst authorLink
10. Improving BLSTM-Based Particulate-Matter Prediction Using Observations from Neighboring Regions
    인근지역 관측데이터를 활용한 BLSTM 기반 미세먼지 예측 성능 향상에 관하여
    Seongyeop Jeong, Inyoung Park, Hong Kook Kim
    KICS Summer Conference 2019
    PaperFirst authorLink

Patents

• Temperature Prediction System
  H. K. Kim, S. Jeong, I. Park · GIST
  US 11,754,748 B2 · KR 10-2875113 — Granted (US 2023, KR 2025)
• System for Customizing an AI Model Based on User Feedback, and Method Thereof
  B. Yoo, S. Jeong · INEEJI
  KR 10-2873944 · WO 2026/142385 — Granted (KR 2025); PCT filed
• Method and System for Commodity Price Prediction Using Dynamic Reconfiguration of Multiple Indicators
  B. Yoo, S. Jeong · INEEJI
  KR 10-2026-0132935 (pub.) · WO 2026/182586 — KR application published; PCT filed
• Electronic Device Including a Text-to-Speech (TTS) Model and Control Method Thereof
  J. Sung, S. Jeong, J. Choi · Samsung Electronics & KAIST
  US 2024/0161747 A1 · WO 2024/106934 · KR 10-2024-0071965 — Published (pending)
• Automatic Gas-Leak Inspection System for Life Rafts Using a User Terminal
  J. Shim, G. Yoon, B. Lee, S. Jeong, D. Kim · Korea Maritime & Ocean University
  KR 10-2080404 — Granted (KR 2020)

Experience

• Ph.D. ResearcherOct 2026 – Present
  KAIST, Kim Jaechul Graduate School of AI — SAIL · Seoul, Korea
  • Explainable AI for time-series and weather-forecasting models.
  • Member of the lab seminar organizing team.
• Senior ResearcherJun 2023 – Oct 2026
  INEEJI Corp. · Seongnam, Korea
  • Led AI R&D for national weather-AI projects with the National Institute of Meteorological Sciences (NIMS): NWP forecast-error prediction (sub-project lead), radar–gauge precipitation correction and nowcasting for high-impact weather, and a Korean weather/climate foundation model (data and training infrastructure).
  • Designed and implemented the company's XAI library (attribution, PDP/SHAP, time-series support) used in its explainable-prediction platform.
  • Industrial forecasting PoCs (steel, chemicals); inventor on three patent filings.
• Ph.D. ResearcherMar 2021 – May 2023
  KAIST, Kim Jaechul Graduate School of AI — SAIL · Seoul, Korea
  • XAI-based automatic correction of artifacts in neural text-to-speech (industry–academia project with Samsung Electronics; IEEE TASLP 2025, first author; patent).
  • Explainable-AI interface for a precipitation nowcasting model with NIMS (HCII 2023).

Education

• Ph.D. student in Artificial IntelligenceMar 2021 – Present
  KAIST, Kim Jaechul Graduate School of AI
  Statistical Artificial Intelligence Lab (SAIL), advisor: Prof. Jaesik Choi
  On leave 2023–2026 for industrial research at INEEJI; returned Oct 2026
• M.S. in Electrical Engineering and Computer ScienceMar 2019 – Feb 2021
  Gwangju Institute of Science and Technology (GIST)
  Advisor: Prof. Hong Kook Kim
  Thesis: LSTM-based Verbal Aggression Detection Using Audio and Recognized Text
• B.S. in Electronics and Communication EngineeringMar 2015 – Feb 2019
  Korea Maritime & Ocean University

Honors & Awards

• Best Paper Award, Korea Software Congress (KSC 2023), KIISEDec 2023
  “Automated Correction of Artifacts in Neural Text-to-Speech Models”
• Appreciation Plaque, Samsung Electronics DX Division (2023)Aug 2023
  For an industry–academia project
• DACON Paldang Dam Water-Level Prediction AI CompetitionSep 2022
  Top 1.5% (public) / top 2.5% (private leaderboard)
• 3rd Place, 2020 AI Grand Challenge (4th round), Track 2: Speech Recognition for Threat Detection2020
  Ministry of Science and ICT, Korea · team WisenetAI-GIST

Invited Talks & Presentations

• Invited talk — “Precipitation Forecasting and Numerical Weather Prediction Error Estimation Using AI”Dec 2025
  2025 KAIST XAI Tutorial Series (Industrial Applications) · Link
• “Recent XAI Research in Speech Domain”Aug 2025
  KAIST SAIL Workshop 2025
• Oral presentation — “Inference-Explanation Techniques and Explanation Display Methods for an XAI-Based Weather Prediction Model”Apr 2023
  KMS Spring Conference 2023, NIMS Outstanding R&D Results Special Session
• Tutorial — “Explanation Tools of Google What-If & LIT”Feb 2023
  2023 KAIST XAI Tutorial Series · Link

Academic Service

• Reviewer — paper review and AI poster judgingNov 2023
  Korea Software Congress (KSC 2023), KIISE
• Organizing staff — data challenge (EDA, baselines, Kaggle competition setup and leaderboard validation)Sep 2021
  PHM Asia Pacific 2021 Data Challenge · Link

Selected Projects

• AI data fusion for high-impact weatherApr 2025 – Oct 2026
  NIMS · KAIST-led consortium, INEEJI

  Extended an operational nowcasting model to East Asia with multi-node GPU training, and built a deep-learning grid-scale radar–rain-gauge precipitation correction (warning-event CSI 0.472 → 0.525). ACM 2026 poster.

• AI prediction of numerical-weather-prediction forecast errorsMay 2025 – Oct 2026
  National Institute of Meteorological Sciences (NIMS) · INEEJI (sub-project lead)

  Six-year KIM model dataset (19 TB), a testbed of eight forecasting architectures (CNN, ConvLSTM, ViT, PatchTST, …) for 500-hPa geopotential-height and 850-hPa temperature error prediction, and precipitation-model recommendation. AGU26 oral.

• Korean AI weather and climate foundation modelApr 2026 – Oct 2026
  NIMS · KAIST-led consortium, INEEJI

  Prototype of the data and training-operations infrastructure for a weather/climate foundation model.

• XAI library for industrial prediction modelsAug 2023 – Jun 2024
  INEEJI

  Attribution (Captum-based), partial-dependence and Shapley analyses with time-series support, packaged with Docker and integrated into the company's cloud explainable-prediction platform; used in industrial forecasting PoCs.

• Explainable-AI-based correction of artifacts in neural text-to-speech2022 – 2023
  Industry–academia project with Samsung Electronics · KAIST SAIL · Project page

  Detects erroneous synthesis segments, identifies the responsible units with layer-wise relevance propagation (LRP) and corrects them by adjusting their weights — reducing attention-alignment errors by 25.86% (CMOS +1.34). IEEE TASLP 2025 (first author); patent applications in KR/US/PCT.

• Explainable-AI interface for precipitation nowcastingApr 2022 – Apr 2023
  Korea Meteorological Administration / NIMS · KAIST SAIL

  Forecaster requirement study (16 online and 3 in-person sessions) and an XAI interface for a radar-based U-Net nowcasting model. HCI International 2023.

• NWP-based temperature and particulate-matter prediction; verbal aggression detection2019 – 2020
  GIST (M.S.)

  Bi-LSTM/CNN/attention models combining regional NWP forecasts with observations (Sensors 2021; US/KR patents), and audio–text verbal-aggression detection (M.S. thesis).